Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Networking Hardware

The Road to 100 Gigabit Ethernet 109

darthcamaro writes "InternetNews is reporting that a grassroots effort is being formed to push 100 Gigabit Ethernet into the mainstream. That's 10x faster than the current fastest Ethernet standard 10 GbE and 1000 times faster than "FastEthernet" but it's not going to be here anytime soon. From the article: '"A group of companies have formed to approach the IEEE to get a vote within the IEEE body to start a standard and that's really where we are," Garrison told internetnews.com. [...] The process then to becoming a full standard is a long and drawn out one that could take five or more years. Garrison explained that the first part of the standard will look at technical and economic feasibility, as well as LAN and WAN opportunities.'"
This discussion has been archived. No new comments can be posted.

The Road to 100 Gigabit Ethernet

Comments Filter:
  • by 0110011001110101 ( 881374 ) on Wednesday February 08, 2006 @02:23PM (#14671206) Journal
    FTFA - Don't expect to see 100 GbE anytime soon, though; it's likely not to see the light of day until at least 2010.

    Sweeeeet! Just in time for me to kick ass and chew bubblegum in super high speed on Duke Nukem Forever...

    To be sure I'm first in line, I'll take my flying car and digital Paper directions. I'm sooo gonna get laid.

    • Come on folks, it's really a funny joke. Don't y'all "futurists" get your panties in a bunch over somebody poking fun at you. First it's modding down, next thing you know slashbots are firebombing embassies. All this over a little satire.
  • Even PCI-X? I'm sure, these will improve too, in the future, of course...
    • It sounds like we just need a PCIe x400 standard. At 2.5Gbs per link times 400 links is a nice 1000Gbs.
    • Wider? (Score:1, Informative)

      by Anonymous Coward
      I realise your comment has been moderated "insightful" but could you explain it for us slow people.
      What does 100G ethernet have to do with internal bus width?
      Thank you.
      • Re:Wider? (Score:2, Informative)

        by mlheur ( 212082 )
        If your PCI bus can only serve data to your NIC at 50Gbps, your card that can send at 100Gbps won't have anything to send that fast because he's waiting for the computer to feed him.

        Now of course that goes away if you only use 100Gbps as trunks between switches, and connections to individual PCs stays at GigE, but the internal bus of the switch still comes into play.
        • What's the current speed that our hard drives are able to do, symmetrically? If we don't have the ability to access and transmit our data that fast at the time, what's the need?
          • Well, just like the grandparent indicated, this is a standard being developed solely for the datacenter, and most likely only for trunk lines between datacenters.

            Companies like ethernet because it is scalable. The same 10Gb trunk lines can, through a simple switch, talk to 1Gb data closets, which can talk to 100Mb clients. The scalability is key because copper is cheap, but doesn't run well at long distances above 1Gb. Faster ethernet standards such as 10Gb ethernet are limited to 15m for copper (10GBASE
      • His thought is that a PCI-X slot can barly send 1Gb so the talk about 100Gb is silly. Of course these 100Gb connections arn't for mortals anyways.
        • Re:Wider? (Score:1, Informative)

          by Anonymous Coward
          Ok, I get it now.
          Actually I work for a router company. So I look at 100G as an interconnect between routers not as a termination point.

          BTW 40Gbe is the most likely next interface. But some ISPs have an all ethernet network (no sonet) so 100G is the next logical step to them.
          Sonet based shops have always been more reasonable; 4x (oc12->oc48->oc192) speed each generation not 10x (10->100->1Gbe-10Gbe).

          OC192 and 10Gbe are the same rate so there is lower cost because of shared components (optics/f

      • "Wider", you know, as in "bandwidth"...
    • I would use this to connect 10Ge switches together, not to connect directly to individual servers in the network. Three 32-port 100Ge switches, linked to 32 32-port 10Ge switches by three links (ie, so each 10Ge switch had three links to different 100Ge switches and one link to each of 29 hosts) would allow any two hosts in a 928-node cluster to communicate at full 10Ge capacity (or, more likely, communicate with any number of hosts in that cluster at 10Ge aggregate capacity).

      Something like this: 928 nod [ciar.org]

    • Forget PCI-X and PCIe. 100Gbit Ethernet is faster than your computer's memory bus.
      • 100 / 8 = 12.5

        Dual channel DDR2 at 1066 mhz can hit 6.4 Gbps. More and more systems are coming with a complete dual-channel hookup per cpu-core (a few with more), so for a four way K8 system thats twice the bandwidth of 100 gigabit ethernet. Of course, K8's dont run DDR2, when they do I doubt we'll see 1066 for a while, and DDR2 has such pathetic latencies (5-10-15-20 latencies anyone?) you'd be lucky to get five eights the theoretical. With four way though, thats still more than 100gigabit could manag
      • Workstations won't be using 100Gbit ethernet anytime soon, but 100Gbit isn't that uncommon speed in networks.
        The company I work for has several links running at 140Gbit, though that speed is achieved by DWDM, and isn't raw ethernet.
        Fastest core routers can already operate at terabyte speeds, so once the backbone networks slowly upgrade to support faster rates, expect availability of faster and affordable consumer connections to rise aswell.
  • 2010, it'll be just in time to be too slow.
  • Having a 100Gb home network will begin a Second Odyssey [wikipedia.org] of user frustration: the broadband connection out of the wall will still only be 10Mb.
  • I'll take a 10GBASE-CX4 starter kit for my home network...

    One 4 or 6 port Internet router, 2 PCI-X or PCI-E cards and 2 CX4 cables.

    It's worth 700 euros... max.

    Fuck, just typing this is making me bitter.
  • How about we get 1gigabit to become standard first... i have yet to connect anywhere (home, office, school) with a gigabit connection...
    • Gigabit switches are pretty cheap these days. I rewired my home network, 7 machines with a gigabit switched network. I paid about $80 for a 8 ports gigabit.

      My windows machines are loving accessing my file server on the network now! Though now it is getting time to buy a 16 ports switch and to upgrade the wireless network too. My 11b connection is getting too slow to work on the laptops...

      I probably would have no need for a 100gigabit network at home... not yet... but I know some a company that were maxing o
      • Gigabit yes, NAS no. The only reason you need GigE is if you're pulling lots of files over the local network. I run myself a Linux box with 4 HDDs over GigE. Cheap, but not a very consumer-friendly solution. Something like a Thecus 4100 (4 SATA drives) is $800. Personally I think the prices for what is essentially a NIC, RAID chip and cabinet are completely off the scale.
    • In a way, you may be connecting through 1GB and just not know it. At the networks here, we have a 1GB connection from the server to the switch, when then splits into 10/100 connections. This means that multiple network machines can grab data in the 10/100 range, but the overall data consumption from machines connected to the server will not be bottlenecked at 100MB/s.

      Of course, depending on what is being done the data rate is still limited by the hardware on the server (drive speeds, etc), but it does hav
      • We're beginning to move to gigabit at my office as well. From switch to switch, it's fiber. Most of our drops are gigabit now, and any new drops are of course gigabit. The goal here is to eliminate bottlenecks, and even though some switches are 10/100, others are gigabit and when they exchange data with each other the bottleneck is not the wire.

        It's also beneficial using gigabit just to keep total network saturation lower. You can max out a 10/100 connection with enough clients pulling enough data f
    • I put Gigabit in a (private) school two years ago. At the time it was still too expensive for their limited budget to make it 100% gigabit on good managed switches, but we were able to put in enough for a minimum of one GbE port in each room of the building. Many of the rooms have more then one GbE port. At the same time we upgraded the backbone of their network (including the older buildings) to GbE so at least the switches have good bandwidth to each other everywhere. They haven't utilized the potenti
    • True. However, GigE is used a lot on backbone connections within LANs. So even though your client is only seeing a 100Mb port, the switch it's plugged into probably uses 1Gb on its uplink side.

      I think the reason you don't see more 1Gb client ports, is because most network architects want to have the backbones be some multiple of the client-link speeds, and 10Gb connections are cost-prohibitive. And if you don't have the bandwidth on the backbone to carry the traffic (and if most of the traffic in your LAN i
  • so what? (Score:4, Interesting)

    by pair-a-noyd ( 594371 ) on Wednesday February 08, 2006 @02:39PM (#14671385)
    This won't help anyone, what with the plans to throttle services.
    All the talk about multi-tiered service and restricting/blocking content is heating up. Who will benefit from this? Only the few that can afford to shell out for premium services. Us little people will all end up with dial up grade service despite the fact that we COULD have better, provided we are willing to mortgage our homes and sell our souls for better speeds.

    I hope the people drag the scumbag parasite profiteers out of their ivory towers and burn them at the stake.

    Will we ever realize the full benefit of high speed Internet? Doubtful. It will be priced out of range of mortals..

    • No, it's worse. We'll pay for mega-expensive outside connections at 50 Mbps speeds, and have nothing that the ISPs will allow to be fast enough to use it all. We'll have to be downloading a dozen things at once to saturate the connection. But that isn't strictly speaking relevant to this. This relates to a intranet connection, like within an office. The same principle applies: there is nothing coming that'll use this kind of speed. Unless you stream dozens of movies over one cable, it's currently poi
    • Re:so what? (Score:3, Insightful)

      by PylonHead ( 61401 )
      Ethernet is a local network protocol, and doesn't have much to do with the way you communicate across the internet. This will help anyone running more than one machine in a particular location.

      Businesses will benefit, and users with more than one machine at home will benefit.

      • Ethernet is a local network protocol, and doesn't have much to do with the way you communicate across the internet. This will help anyone running more than one machine in a particular location.

        This you say, but look at networks in Hong Kong [hkbn.net] & many European cities. Or Wellington, NZ [citylink.co.nz], where I live and work. Metro Area Ethernet [wikipedia.org] which connect hundreds of buildings - including residential blocks.

        The future of delivery to Multi-tenant buildings (apartment & office blocks) is all about Ethernet. Anything b
        • That's all fine and dandy if you live in an advanced country like Hong Kong, Australia/NZ, or someplace in Europe. But for those of us stuck in the USA, we're going to be limited to DSL or cable, and in the very near future our bandwidth will be severely limited due to the profiteering of greedy American telcos who own the backbones.
    • You'd need entire google datacenters to layer7 packet shape a 100 gigabit stream.

      Here's to hoping our bandwidth exceeds moore's ability to packet shape it!
    • One simple and powerful application for even higher bandwidth ethernet is iSCSI and SANs.

      For enterprises, its a no-brainer.

      For the consumer, eventually prices will come down and it would really be cool to be able to put all your storage in one location in a well designed RAID with hotswappable drives and expandable storage and access it like you access any block device (ie. install OSes on it, etc.).
  • by Guspaz ( 556486 ) on Wednesday February 08, 2006 @02:43PM (#14671432)
    Those terms imply consumer acceptance. Even the fastest consumer hard drives can't saturate a 1 gigabit ethernet connection. Consumers don't even need 10 gigabit, why would they want 100 gigabit?

    Besides, while 1 gigabit ethernet has gained consumer acceptance over the years, with more and more consumer-level products supporting it, the vast majority of consumer networks are still 100 megabit. Most new computers might have onboard gigabit ethernet, but since manufacturers keep putting 100 megabit switches in convergence products (routers with onboard switches), nobody can use gigabit.

    Of course, I realize that the article uses these terms in relation to large companies, but I don't think they can be used in that context. Even so, the current equipment to handle 10 gigabit connections is quite expensive even for large corporations, the cost of 100 gigabit would be prohibitive.
    • If you want to make sure you have 24 vlans in an wide-lan to have 1 gbps each then it's likely cheaper to have multi-gbps cards and cables.
    • Even the fastest consumer hard drives can't saturate a 1 gigabit ethernet connection.

      My keyboard can't saturate a 1 gigabit ethernet connection, either. A nonsensical observation, but no more than yours.
      • Of what good is a 1Gb/s ethernet connection to a consumer if his hard drive can only read/write data at a peak rate of 400 or 500 Mb/s ? The speed of the LAN is irrelevant at that point.

        • client$ nc -l -p 5000 > /dev/null
          server$ cat /dev/urandom | nc client 5000

          Some of us like to transfer data that never hits the disk.
        • If the data is stored in Ram, and there are multiple threads working, you could saturate the connection. True, that you will likely just be saturating the network switch more often though.
          • So the "mainstream" needs 100 GBit Ethernet to move data from RAM of one computer to the RAM of another computer?

            100 GBit/sec - that's about 10 GByte/sec. You'll need a 64 bit CPU to even saturate the connection for more than 0.5 seconds.

        • With larger and larger storage needs (for such things as DVR, centralized content management systems, music siloing etc...) and for interactivity between more and more devices on the network, I can certainly see a point at which 100mbs will not suffice for most users.

          Over the next 5 to 10 years we will see an explosion of applications, and NAS (network attached storage) devices and multiple computers and computerized devices will enter the home network and become the norm.

          It is coming; lead, follow, or get
        • That's easy. Joe Schmo's fast ethernet NIC maxes out at 90Mb/s with TCP/IP overhead. Jow Schmo's new hard drive maxes out at 500Mb/s sustained read speed.

          THATS A 5.5x INCREASE IN PERFORMANCE just going gigabit over fast ethernet. Sure, you're not saturating the gig pipe, but you're certainly getting your money's worth.

          Visualize it this way: you COULD wait about 15 minutes to transfer a DVD9 over fast ethernet, or you could max out your hard drive over gig ethernet and take only 3 minutes.

          And when you co
        • Of what good is a 1Gb/s ethernet connection to a consumer if his hard drive can only read/write data at a peak rate of 400 or 500 Mb/s ? The speed of the LAN is irrelevant at that point.

          There's this thing called "disk caching." Maybe you've heard of it. On servers these days there are many cases where pretty much everything you'll ever want to serve off the disks can be cached in memory, which totally removes disk I/O speed as a bottleneck.

          Also, you're assuming that everything I ever might want to serve is

    • The point isn't to run this out to the desktop, or server. Even 10gigE is a silly choice for that today. But 10+gigE is plenty useful for aggregating between links. For example, if my company has two branch offices and I need to transfer a lot of data back and forth, I might lease dark fiber and connect the two offices using 10gigE between each office's routers.
      • And my previous point stands. Only the largest of companies can currently afford 10GigE optical equipment. ThePlanet, a large hosting company with several dozen gigabits of bandwidth has been wanting to move some of their connections from GigE over to 10GigE for a while. Unfortunately, the cost and availability of the equipment to handle 10 gigabits per second has been a stumbling block.
      • Huh? With external storage (SANs), it's easy to saturate a 1 GbE connection. Pretty soon, 10GbE will be pretty normal if you want a decent-performance SAN.
    • So you suggest we all stick to our 28.8kbps modems, on our 486s with 8MB of RAM, because "the cost of anything higher is prohibitive". Just because it *is* {unusable, unsupportable, too expensive, uneconomic to produce, ...}, doesn't mean it wont be (in the near future)
      • I don't think you fully understand the challenge of constructing 100 Gb/s equipment. The optical industry is struggling with developing 40Gb/s at a competitive price. No telco would ever touch that if you can get 4 x 10Gb/s equipment at a fragment of the cost.

        Cost is not necessarily prohibitive, but it certainly can slow down acceptance of new technology.

        Besides, we should focus on higher bandwith connections to the end user such as fiber-to-the-home initiatives. This is one of the few things that can
    • by Meostro ( 788797 ) on Wednesday February 08, 2006 @04:28PM (#14672378) Homepage Journal
      "I think there is a world market for maybe five computers." - Thomas Watson, 1943

      "640k ought to be enough for anybody" - Bill Gates, 1981

      "Consumers don't even need 10 gigabit, why would they want 100 gigabit?" - Guspaz, 2006
      • "I think there is a world market for maybe five computers." - Thomas Watson, 1943

        - Yes, at the time. I know I wasn't buying one in 1943. Were you?

        "640k ought to be enough for anybody" - Bill Gates, 1981

        - Yes, at the time. I know I couldn't afford more that that in 1981. Could you? Secondly, I don't believe he ever said this.

        "Consumers don't even need 10 gigabit, why would they want 100 gigabit?" - Guspaz, 2006

        - Yes, even in 2010 consumers won't need 10 Gig Ethernet. Step outside your living room and realize
    • Raptor SATA can break 88 MBps. Thats 704 Gbps. Add a little bit of network overhead, even for something as raw-metal as ATA-over-IP, you're still talking "damned close". Thats for _one_ hard drive, how passe.

      Of course 10 gb is not for "consumers". At least not now, not for what we've got. But some day, if homes really are shuffling around 1080p streams (or bigger) like candy, even your average consumer home might need 10 gb backbones. If 1 gigabit is one single modern hard drive, 10 gigabit is not unr
      • Raptors are bad ass. I have two of the 36gb version in a RAID0 config. I can sustain over 110MB/sec. At a LAN party, I moved all my "tradable files" to this array and plugged in to the Gig-E switch with my NForce3 Gig-E adaptor.

        That link was saturated the entire night.
        • supposedly the new ones also dont sound like jet engines.

          the original 36 gig ones were strange beasts indeed. very quirky. sometimes they worked great though. great for doing fileserve. just make sure to reserve some bandwidth for yourself, eh? ;-]
      • Raptor SATA can break 88 MBps. Thats 704 Gbps. Add a little bit of network overhead, even for something as raw-metal as ATA-over-IP, you're still talking "damned close". Thats for _one_ hard drive, how passe.

        I think you mean 0.704 Gbps. I don't know about gigabit NICs, but with 100mbit NICs I usually see about 80mbit of real-world transfer speed using regular Windows filesharing. Still a bit of leeway there, if the percentage holds true.

        Besides, that 88MB/s figure is the peak sequential transfer (at the edg
        • right now, i agree with you 100%, 10g is hardly needed anyhwere. but to address your orignal post, its that future thing that has me wondering. doubling up or tripling up on gigabit is currently standard practice for data intensive systems-- not just big iron, but rouge hackers cobbling together powerful clusters and grids for fun or profit. and it makes sense, its so damned cheap why not? besides, who can afford 10g? but with the throughput wars, how long can this really last? this is grassroots. th
    • Technology like this is not aimed at home users...
      Nor was gigabit, or 100mb ethernet when it first became available, it took a few years.

      For a while, 100mb ethernet will be used in datacenters between high performance multiple processor servers connected to large fibre channel sans... not cheap consumer hardware
  • Jumping the Gun (Score:3, Insightful)

    by Nom du Keyboard ( 633989 ) on Wednesday February 08, 2006 @02:44PM (#14671440)
    but it's not going to be here anytime soon.

    So why are we even talking about it now? This isn't going to change anybody's life (unless you've trying to get on the standards committee) today, tomorrow, or likely this year. How about this be reopened when some working silicon (or whatever material it's going to take to operate at this speed) is up and working in the lab? Then it might have some relevance.

    • So we shouldn't talk about it until it's a viable product, but how does it become a viable product without any sort of discussion or planning? Good thing our ancestors had a different mindset, or we'd still be living in caves.
  • Pointless (Score:1, Redundant)

    by Ramble ( 940291 )
    There isn't really much point, very few internet connections are going to be over 10mbit/s let alone 1000gbit/s. The only application this is really going to have is transferring huge files, and even then, hard disks arn't fast enough to saturate the network bandwidth.
    • Re:Pointless (Score:3, Insightful)

      by TinyManCan ( 580322 )
      10 Gigabit Ethernet can be effectively used. I've seen transfer rates of over 850 MB/s on a single 10gbs link.

      You might be right in that a single consumer drive can not make use of that storage, but there are systems out there that can saturate a 10gbs link many times over.

      Just because you can not fathom a use for the technology, does not make it pointless. Just try managing an environment with 50+ backup servers (because of the 1gbs and 100mbit links to those servers) compared to an environment that ha

    • I can burn 1.5 Gbps with a single HD video stream. Network HD video feeds are distributed at 45 Mbps.
  • ...adding a fiber adapter to motherboards as standard? With the limitations that wire has, is a fiber connection directly on your motherboard, or as a cheap alternative add-on card, that far off?

    Verizon already offers Fiber To The Home [verizon.com] in some markets. Imagine a direct fiber connection to your PC.
      • Fiber Channel only runs at 2Gb/sec, which sucks. With 10Gb ethernet pushing its way out and FC not reallydoing anything new (though admittedly I could be out of the loop on this), I'm wondering if FC will be replaced by ATA-over-ethernet for SAN usage.
        • "Fiber Channel only runs at 2Gb/sec"

          Currently seen:

          10Gb/s ISL

          4Gb/s host and storage.

          You can also aggregate fibres and get more bandwidth.
          • Also, FC is faster (in terms of payload bytes) than iSCSI at the same wire speed, because of overhead.

            I haven't seen a standard for ATA over ethernet, but there is plenty of storage today that uses iSCSI to talk to ATA drives. iSCSI has a lot of overhead compared to fibre channel, however. 10Gb iSCSI will be faster than 4Gb FC, but not by much, and only if everything is switched so collisions are rare (which is the norm these days). 10Gb FC is significantly faster than 10Gb iSCSI in terms of payload tran
          • Ahh yes. I stand corrected. I googled this and found a 2001 press release from Brocade saying they were releasing 10Gb FC switches. I'm a bit out of the loop on this.
    • Sun and Luxtera are working on it [luxtera.com]. It won't be long before it's possible. Their method sounds inexpensive as well.
    • I've got a 15mbit connection at home, and its rare I see a transfer get that high. Once in a while I'll get an average speed around 1.1mb/sec but to max out the connection I have to hit a very unused server that is very close to me.

      Until the places you want to connect to can handle that sort of load, wired network is just fine. Considering my house is all gigabit without fiber, it'll be a LONG time before the connection out of my house outstrips the speed of the network inside.
    • The problem with fibre to the desktop is that it is so expensive to terminate and too fragile to put behind Joe-User's computer. When he starts moving his cables around and bends the fibre, the fibre cable will break internally and will need a replacement. These replacement fibre cables are not cheap, and are far more expensive than Cat6 cables.

      The best method of incorporation is simply to have a fibre backbone to the horizontal cross-connect of the facility (Or the local Switch in a small environment), and
    • You mean like this motherboard [supermicro.com] with four copper, plus four SFPs that can accept copper or fiber modules? Now if only they could combine it to 2 10 Gig Ethernet plus two Gig Ethernet, that would be one kick-ass server :)
    • Well, it's not cheap, but my school uses fiber directly to PCs.
      Case Western Reserve University, (we've been on slashdot [slashdot.org] before) home of Fiber Gigabit Ethernet to the student systems.

      We even have a gigabit link to the internet these days.
      It's very nice to get sustained internal transfers of 25 MB/s or so...
  • have you even looked at the signal integrity issues related to any hss link running at 10gbps/link? unless you run 10 lvds hss for half-duplex 100gbps it'll have to be optical. then it becomes an issue of designing an optical switch that can handle the load, and a ridiculously (and impossibly so by today's standard) fast optical-to-electrical interface, again, to at least 10 hss lvds pairs to achieve those speeds... dream on!
  • by 0rbit4l ( 669001 ) on Wednesday February 08, 2006 @03:04PM (#14671630)
    100 Gb/s Ethernet is not for joe schmoe sitting at home on his consumer broadband connection - it's for servers connected to a backbone link. The idea is that 100 Gb/s Ethernet would be part of a server that could handle far more connections & deliver far more throughput per machine, which is important in the datacenter where you'd like to reduce the raw number of machines to save on power (and as part of that, cooling). Sheesh, if you'd stop bitching for a few minutes about your cable company limiting the rate at which you copy donkey porn, you might discover a whole other world of networking challenges out there that people are working on.

    (Obviously you have to have enough bus & memory bandwidth and compute power to drive a 100 Gb/s link - but this is a necessary piece of the puzzle).

    • This is just what I was going to post. Around here, it seems very few can relate to technology that doesn't make sense to a home PC.

      Faster network speeds can only improve the computing state of the art, as tech from big hardware trickles down to stuff we can afford to buy ourselves after a time.
    • I don't think anyone is missing the point...I agree that there is a sub-section of users that are bitching about the capping of download/bandwidth, however, these same people are actually fighting (whether they know it or not) against another few evils. They are also, I'm sorry to say, the reason there is an Internet...without end-users, albeit corporate/commercial/personal, what's the point of an Internet?

      Remember when cell phones were per-second? All plans eventually switched to per-minute, why? ~30% L
    • Buddy, YOU are missing the point.

      It's for MAN/WAN/LAN switch-switch or router-router connections. Not host to switch connections. Why do you think you can't have OC48/192/768 on a NIC...

    • Just think how much more of the InterNet the NSA can tap with these!
    • by jd ( 1658 )
      I was hoping there was a chance of 100 gigabits to the home, or at least the one gigabit the Japanese get to their homes. Who cares what hard disks can handle, if you're using streamed media, Grid Computing or just an old-fashioned RAM disk? (Besides, if you use multiple hard drives in a striped array, you can get far higher throughput than from a single drive.)
    • 100 Gb/s Ethernet is not for joe schmoe sitting at home on his consumer broadband connection - it's for servers connected to a backbone link.

      Apparently, you missed the point as well. Servers today seldom push even 1Gb Ethernet, let alone 10Gb or 100Gb Ethernet. The primary use for 10Gb Ethernet today is in the network backbone - Comcast, for example, has an entire fiber network that carries voice and data based on 10Gb Ethernet.

      While we may see server adoption in the future, it's going to be quite a while b
  • You can only twist and shield the wires so much before you can't twist them anymore and extra shielding does jack. They've got to be hitting the limits of twisted pair copper wiring. It's gonna be a bitch to get it up to 100Gb, and it's going to be even harder (if it's even possible) to get that up to a terabit.
    • They gave up on UTP cabling for the 10gbe standard. 100gbe will be fibre.
      • They gave up on UTP cabling for the 10gbe standard.

        No, they just don't have a standard out yet; 802.3an is working on it though. It wasn't released simultaneously with the fibre standard because it's a harder problem. Then again, neither was gigabit. Or come to that even 100Mbit/s, depending exactly how you date it.

        100gbe will be fibre.

        Probably.
    • We could also figure out how to do DWDM or other technologies that are easy with radio to put data on the wire in a more effecient manner. Going to fiber works, of course, but then you still have the problem with switches and GBICs and the like: the signal has to be in copper at some point, and 100Gb pushes the limits on that until we start using better encoding.
  • just out of interest, will this 100gigabit networks run on Cat6 or Fibre??

    could anyone please enlighten me, i couldnt find much info, thanks
    • Re:confused (Score:3, Informative)

      by Bananatree3 ( 872975 ) *

      Most likely 100Gbit will run on Fiber Cables, which have MUCH more bandwidth than their copper-based equivelents (Cat6,etc.)Wikipedia says that they have the potential to carry terabits of data per second [wikipedia.org].

      For the ethernet cables, according to wikipedia, Cat6 is reliable up to 1Gbit connections [wikipedia.org]. However, Category 7 [wikipedia.org] cables have been developed for 10Gbit connections. It seems to me that it might be possible to push ethernet cables up to 100Gbit. But that is a BIG if, as I don't know how much further the st

    • It might run on copper but if it does it will probably be parallel.

      You can run 10 gig ethernet over copper for a few inches. You can run it about 45 feet if you channelize it like they do with CX4. It's four lanes of copper, each 3 and 1/8 gigabits a second.

      The next Ethernet is more likely going to be 40 gigabits. They will just take the 802.11ae Clause 49 PCS, which is a serialized 64 to 66 bit data stream, and multiply it by 4 to get 4 times the throughput. Google around for 40 gig ethernet and you wi
    • I would guess they would use fiber, since copper just doesn't have that much data capacity (maybe with coax it could be done, but at any rate it would likely require expensive cabling and only work for short distances). Fiber is much better for this sort of thing.

      A more interesting question is whether to use singlemode or multimode fiber, if they go that route. Most "normal" lan hardware uses multimode, which in general is good for connections of tens of gigabits over distances of 2km or less. Singlemo

      • I would guess they would use fiber, since copper just doesn't have that much data capacity (maybe with coax it could be done, but at any rate it would likely require expensive cabling and only work for short distances)

        And just what metal do you think coax cables are made from, hrm?
        • And just what metal do you think coax cables are made from, hrm?

          I wasn't trying to imply that coax was not made of copper, though I can see how it may appear that I was. I am sorry for the confusion, I was merely trying to indicate the conditions under which copper might work. At any rate, I don't think connecting servers together with, say, garden-hose sized LMR-600 cables and nics with N-connectors instead of RJ-45 jacks is going to be the wave of the future, though it is amusing to contemplate. (Doe

  • I used to work in the InfiniBand space where folks are using host adapters at 20 Gbit (4X, Double Data Rate). Some of the big server vendors are doing 30 Gb (12X Single Data Rate) host adapters. With all of this host speed it is only a matter of time before the switch to switch links will go up in speed.
    High speed systems like this are getting used in high performance computing to build larger clusters. Having faster switch links will allow these fabrics to be created with less switches and thus less hop
  • Just a random thought that popped up in my mind: The speed of light is 3*10^8 m/s, and you want to send a bitstream of 10^11 bit/s. If we forget that the actual speed of an electrical pulse is lower than the speed of light, this results in a mean travellenght of 0.003m, or 3 millimeter per bit. So in my utp-cable I will have a bit travelling through it every few millimeters. Isn't this going to cause severe trouble when e.g. one of the eight wires is a bit longer than the rest? Or how tight will you need to
  • Current networking technologies are already a lot faster than what is commercially available from the phone and cable companies. It's not going to help you much if you have a 100Gbit NIC but only a 5Mbit internet connection. You're really only going to see this benefit you if you play a lot of games over a LAN or for businesses that need a high bandwidth LAN, IMHO.

No amount of careful planning will ever replace dumb luck.

Working...